Weidhaas Joanne B, McGreevy Kristen M, Marco Nicholas, Sundahl Nora, Cabanski Christopher R, Spencer Christine, LaVallee Theresa, Ost Piet, Telesca Donatello
UCLA David Geffen School of Medicine, Los Angeles, CA, USA.
Department of Biostatistics, UCLA, Los Angeles, CA, USA.
J Transl Med. 2025 Jul 28;23(1):848. doi: 10.1186/s12967-025-06842-3.
Germline microRNA-based variants (mirSNPs) have been shown to be predictive biomarkers of toxicity and tumor response across cancer treatments, including to anti-PD1/PDL1 immune checkpoint therapy. CTLA-4 inhibitors are another immune checkpoint inhibitor with known significant toxicity in the form of immune related adverse events (irAEs). The potential of mirSNPs to predict irAEs and/or response to anti-CTLA-4 therapy alone has not previously been reported and was the purpose of this investigation.
We evaluated genetic signatures to predict toxicity and tumor response to anti-CTLA-4 treatment alone in melanoma patients using three separate cohorts. DNA was extracted from blood samples from 77 patients treated with anti-CTLA-4 therapy and analyzed using a custom panel of mirSNPs. We employed a combination of Elastic Net, Random Forest, and Boosted Tree models, incorporating germline mirSNPs, patient demographics, and treatment variables to predict toxicity in the form of irAEs or disease response. Additionally, we conducted a comparative analysis of gene ontology (GO) pathways to discern biological differences influenced by these genetic markers.
We developed two unique mirSNP signatures predicting toxicity or response to single agent anti-CTLA-4 treatment. These signatures both have excellent predictive accuracy with AUCs of 0.793 for toxicity and of 0.842 for response. The signatures do not overlap, nor is the toxicity signature similar to the toxicity signature for anti-PD1/L1 single agent therapy. Through GO analyses we found that both of these signatures have biological pathways involved in pri-miRNA transcriptional regulation, yet also have unique pathways that differentiate them.
Our findings continue to support the utility of mirSNPs as predictive biomarkers of immune checkpoint therapy, for both toxicity and response. Further investigation in larger, diverse cohorts as well as to dual checkpoint inhibitor treatment is a planned next step to further their application.
基于种系微小RNA的变异（mirSNP）已被证明是癌症治疗（包括抗PD1/PDL1免疫检查点疗法）中毒性和肿瘤反应的预测生物标志物。CTLA-4抑制剂是另一种免疫检查点抑制剂，已知会以免疫相关不良事件（irAE）的形式产生显著毒性。此前尚未报道mirSNP单独预测irAE和/或抗CTLA-4治疗反应的潜力，这也是本研究的目的。
我们使用三个独立队列评估了黑色素瘤患者单独接受抗CTLA-4治疗时的毒性和肿瘤反应的遗传特征。从77例接受抗CTLA-4治疗的患者的血液样本中提取DNA，并使用定制的mirSNP面板进行分析。我们采用了弹性网络、随机森林和提升树模型相结合的方法，纳入种系mirSNP、患者人口统计学特征和治疗变量，以预测irAE形式的毒性或疾病反应。此外，我们对基因本体（GO）通路进行了比较分析，以识别受这些遗传标记影响的生物学差异。
我们开发了两种独特的mirSNP特征，可预测对单药抗CTLA-4治疗的毒性或反应。这些特征均具有出色的预测准确性，毒性的AUC为0.793，反应的AUC为0.842。这些特征不重叠，毒性特征也与抗PD1/L1单药治疗的毒性特征不同。通过GO分析，我们发现这两种特征都有参与初级微小RNA转录调控的生物学通路，但也有使其区分开来的独特通路。
我们的研究结果继续支持mirSNP作为免疫检查点疗法毒性和反应预测生物标志物的实用性。计划下一步在更大、更多样化的队列中以及对双重检查点抑制剂治疗进行进一步研究，以进一步推广其应用。
Zotero 插件